Consulting work
Bürokratt
DESIGN & PROTOTYPESMake the system understandable.
Keep the controls clear.
I work with the team on Bürokratt’s AI back-office workflows, connecting product design and UX with the decisions people need to make. My contribution focuses on analytics, data-source configuration and the information hierarchy that helps operators understand and control a complex system.
Bürokratt is a network of chatbots on public sector institutions’ websites that enables people to obtain information from institutions and to use both public and information services via virtual assistants.
Context & discovery
A conversation for the citizen.
A working system for the team.
Behind a virtual assistant, people need to understand conversations, investigate service performance and configure the information available to the AI. My work focuses on these operator and administrator views.
I worked with the team to make the next decision easier to find: which signal needs investigation, which source is being configured and what still needs review. The design keeps technical distinctions visible within the existing interface and system constraints.
My contribution
I worked with the team to turn AI back-office complexity into workflows an operator could follow: what to understand, where to investigate, what to configure and what still needed review. My contribution connected product reasoning, information hierarchy and hands-on interaction design within Bürokratt’s requirements.
In analytics, I brought conversation volume, waiting time, response quality and routes to investigation into one overview. In database configuration, I separated the connection, selected data, field descriptions and LLM index, keeping index status and flagged items distinct from saving. These are design and prototype contributions within the wider team.
Design & prototypes
Bürokratt / Analytics
See the service. Before opening the history.
I reorganised the overview around the questions a team needs to answer: how conversations are handled, where people wait and how responses are rated. The design brings those signals into one scannable view.
Start with the question.Conversation volume, waiting time and response quality sit together in the overview.A person can orient themselves before opening individual conversations.
Keep comparisons in context.Domain, period and comparison controls frame the same set of signals.The view makes the scope of a figure part of the reading task.
Keep the next step visible.The overview connects to conversations, feedback and advisors.The design makes investigation a next step instead of the only way to understand performance.
Bürokratt / Database configuration
Choose the data. Keep the AI’s scope visible.
I worked on making database configuration understandable without removing control. The design separates the connection, the schema being exposed and the LLM index, so each has a clear place in the workflow.
Visible in the design
Scope. Meaning. Review.
- 01Scope
Selected tables and columns make the configured data scope visible.
- 02Meaning
Descriptions explain the source and its fields in the same working view.
- 03Review
Index status and flagged items remain distinct from the save action.
Make the system understandable.
Keep the decisions visible.
Three decisions behind the work
01Start with the decision.Bring performance signals together so the operator can choose where to investigate.
MY CONTRIBUTIONI shaped the analytics overview around what an operator needs to understand or investigate next. Conversation history became a route into the detail rather than the place to reconstruct the whole picture.
VISIBLE IN THE DESIGNThe domain, date range, performance signals and routes to further investigation sit within one overview, providing context for the next decision.
02Reveal the structure.Give the connection, data scope and AI index distinct places in the workflow.
MY CONTRIBUTIONI clarified the hierarchy of database configuration so the operator could follow the relationship between a source, its schema and the LLM index.
VISIBLE IN THE DESIGNSource selection, field descriptions, index status and review items remain separate parts of the working view.
03Keep the rules visible.Make configuration and review states clear within the system’s constraints.
IMPLEMENTATION CONSTRAINTSDatabase configuration depends on distinct source, index and review states. The interface needs to retain those distinctions while making the workflow easier to follow.
HOW THEY SHAPE THE DESIGNI clarified workflows and interface decisions with the team, keeping selected data, descriptions and flagged items visible without treating the save action as proof that every review issue is resolved.
What this demonstrates
Clarity in the system.
Product reasoning, information hierarchy and interface design working together: helping people understand the system, keep control and find the next useful action.
How I contribute
Judgement within constraints.
I help teams find practical improvements within established design systems, technical dependencies and domain requirements.
